Kataguru Sceptic 35B A3B v2.8.1 BF16 needs about 15.7 GB of memory at its main file (a 14.5 GB file) with an 8K context window, so it fits on a 16 GB GPU. This repository has 3 quants, from 12.9 GB to 16.7 GB. Longer context needs more: at 32K tokens the KV cache alone adds about 2.8 GB.
Kataguru Sceptic 35B A3B v2.8.1 BF16 – a 35.2B multimodal LLM
Kataguru Sceptic 35B A3B v2.8.1 BF16 is a 35.2B parameter model with 3 active parameters in a 254‑expert MoE. It supports reasoning, tools, and vision and was trained on 262144 context tokens. The model is released under the apache‑2.0 license.
Good at
- Multimodal reasoning
- Tool use
- Vision understanding
Limits
- Requires 16‑24 GB VRAM for single‑GPU inference
- Large context length demands high memory
- No official vision support beyond the bundled projector
Best for
- Complex multimodal queries
- Agent‑style tool‑driven tasks
- Large‑context document analysis
Who it’s for. Developers needing a high‑capacity multimodal LLM for single‑GPU setups. The model is suitable for research and production workloads that can accommodate its memory footprint.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| — | 12.9 GB | 14.1 GB | Too big | Get file → | |
| — | 14.5 GB | 15.7 GB | Too big | Get file → | |
| — | 16.7 GB | 17.9 GB | Too big | Get file → |
The GGUF files are 12.87 GB, 14.47 GB, and 16.69 GB. At 8 k context the estimated VRAM usage is 14.1 GB, 15.7 GB, and 17.9 GB respectively. For a 16 GB VRAM GPU use the 12.87 GB file; for 24 GB use the 16.69 GB file. Memory estimates are approximate.
What GPU do you need for Kataguru Sceptic 35B A3B v2.8.1 BF16?
These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.
- 4 GB VRAM: no quant fits at an 8K context.
- 6 GB VRAM: no quant fits at an 8K context.
- 8 GB VRAM: no quant fits at an 8K context.
- 12 GB VRAM: no quant fits at an 8K context.
- 16 GB VRAM: the best quant that fits is the main file (14.5 GB file, about 15.7 GB in use).
- 24 GB VRAM: the best quant that fits is the main file (16.7 GB file, about 17.9 GB in use).
Which Kataguru Sceptic 35B A3B v2.8.1 BF16 quant should you choose?
If memory is tight, the smallest file is the first one in the table at 12.9 GB, with the quality trade-offs that come with fewer bits.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Kataguru Sceptic 35B A3B v2.8.1 BF16
How much VRAM does Kataguru Sceptic 35B A3B v2.8.1 BF16 need?
About 15.7 GB at its main file with an 8K context window. That is the 14.5 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.
Can I run Kataguru Sceptic 35B A3B v2.8.1 BF16 on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 14.1 GB, so you would need a larger GPU or CPU offloading.
What is the file size of Kataguru Sceptic 35B A3B v2.8.1 BF16 GGUF?
It depends on the quant: file is 12.9 GB, file is 14.5 GB, file is 16.7 GB.
What context length does Kataguru Sceptic 35B A3B v2.8.1 BF16 support?
The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 0.7 GB at 8K and 2.8 GB at 32K.
Where can I download Kataguru Sceptic 35B A3B v2.8.1 BF16 GGUF?
The files are on Hugging Face in the kataguru/Kataguru-Sceptic-35B-A3B-v2.8.1-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does Kataguru Sceptic 35B A3B v2.8.1 BF16 use?
Hugging Face lists the license as apache-2.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.
Can I run Kataguru Sceptic 35B on 8 GB VRAM?
No, the smallest GGUF file requires at least 14.1 GB VRAM at 8 k context. A larger VRAM budget is necessary.
Does the model support vision tasks?
Yes, it includes a vision projector (mmproj‑f16) for multimodal inputs.
What license applies to this model?
The model is distributed under the apache‑2.0 license.